Consistent AWS timeouts

[excitoon]

I hereby agree to the terms of the CLA available at: https://yandex.ru/legal/cla/?lang=en

Changelog category (leave one):

• Bug Fix

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

Fixed a bug with unlimited wait for auxiliary AWS requests.

[excitoon]

Related: #21852

[excitoon]

Failed tests are unrelated.

[nikitamikhaylov]

BTW, where was the bug? And why these changes fix it?

[excitoon]

@nikitamikhaylov the bug was in few places (not sure if I found every one), for example in S3CredentialsProviderChain:

                /// EC2MetadataService throttles by delaying the response so the service client should set a large read timeout.
                /// EC2MetadataService delay is in order of seconds so it only make sense to retry after a couple of seconds.
                aws_client_configuration.connectTimeoutMs = 1000;
                aws_client_configuration.requestTimeoutMs = 1000;
                aws_client_configuration.retryStrategy = std::make_shared<Aws::Client::DefaultRetryStrategy>(1, 1000);

                auto ec2_metadata_client = std::make_shared<Aws::Internal::EC2MetadataClient>(aws_client_configuration);
                auto config_loader = std::make_shared<Aws::Config::EC2InstanceProfileConfigLoader>(ec2_metadata_client);

                AddProvider(std::make_shared<Aws::Auth::InstanceProfileCredentialsProvider>(config_loader));
                LOG_INFO(logger, "Added EC2 metadata service credentials provider to the provider chain.");

It uses "default" parameters to set timeouts but in our code we switched to another properties at some moment. Actually, requestTimeoutMs is more appropriate name for this, because it effectively sets timeout on a socket, for single read, and this is not about HTTP at all.

Also, in Aws::Internal::MakeDefaultHttpResourceClientConfiguration() (this code shall not be used right now):

            // EC2MetadataService throttles by delaying the response so the service client should set a large read timeout.
            // EC2MetadataService delay is in order of seconds so it only make sense to retry after a couple of seconds.
            res.connectTimeoutMs = 1000;
            res.requestTimeoutMs = 1000;
            res.retryStrategy = Aws::MakeShared<DefaultRetryStrategy>(RESOURCE_CLIENT_CONFIGURATION_ALLOCATION_TAG, 1, 1000);

            return res;